求解大规模优化问题的正交反向混合差分进化算法

Hybridization differential evolution algorithm of orthogonal crossover and opposition-based learning for large-scale optimization problem

差分进化算法简单高效,然而在求解大规模优化问题时,其求解性能迅速降低。针对该问题,提出一种正交反向差分进化算法。首先,该算法利用正交交叉算子,加强了算法的局部搜索能力。其次,为防止过强的局部搜索使算法陷入早熟收敛,利用反向学习策略调节种群多样性,从而有效地平衡算法的全局和局部搜索能力。利用11个标准测试函数进行实验,并和差分进化算法的四种优秀改进版本进行比较,实验结果表明提出的算法求解精度高、收敛速率快,是一种求解大规模优化问题的有效算法。

Differential evolution is simple and efficient. However,when solving the large-scale optimization problems,the performance decreases rapidly. To overcome this problem,this paper proposed a hybridization differential evolution algorithm of orthogonal crossover and opposition-based learning. In the hybrid algorithm,it used orthogonal crossover to enhance the exploitation ability and adopted opposition-based learning to adjust the diversity of population. Thus it could balance the local and global search ability efficiently. It tested the new algorithm on 11 standard benchmark problems and compared with other four famous variants of differential evolution. The results show that performance of the algorithm is better than those of the compared algorithms in terms of accuracy and speed. Thus,it can be an efficient algorithm for large scale optimization problems.